The mining industry is starting to take interest in Machine Learning and Artificial Intelligence, thanks to big data collected by tech-providers like Newtrax.
Productivity in underground mining has recently seen a significant increase by using sensor data (Internet-of-Things) to react to real-time events that can disrupt the workflow of the mine and cause gaps between a manager’s plan and the actual work completed.
One of the most efficient ways of using this captured data is to feed it to Machine Learning algorithms to reap tangible benefits such as reducing interruption of drilling and haulage, predicting breakdown of mobile equipment, predicting potential bottlenecks in real-time, and providing early notifications of hazards.
What has been established is that Machine Learning algorithms typically use Tensor Processing Units (TPUs) to learn and identify patterns using large datasets, without being explicitly programmed to do so. For example, an Artificial Intelligence bot monitoring all vehicle sensor data can become exponentially better at predicting the moment of an undesired breakdown of the transmission of a truck.
“We learned early on that Machine Learning breakthroughs are achieved through the use of good training datasets, and that algorithms support good data and not the other way around” said Michel Dubois, Vice President QA, Newtrax.
With more than 75 underground mines on all continents, Newtrax sensor-based solutions have captured years of big data on real-time events tied to mining vehicles, personnel and the underground environment. By digitizing underground mines with this sensor-based “nervous system”, Newtrax is equipping its clients with the tools to enable optimization with Machine Learning.
Indeed, mimicking the human brain requires very sophisticated tools. However, Machine Learning algorithms are on their way to becoming IT commodities and the challenge is moving towards better training data prepared by industry-specific business analysts.
Since March 2017, Newtrax Technologies has partnered with Montreal-based Institute for Data Valorisation (IVADO), the largest researcher consortium in Canada in Data Science, Artificial Intelligence and Operations Research. Together, they have spearheaded Machine Learning pilot-projects for underground mines using open-source algorithms. The objective of these projects is to test the quality of existing datasets in order to better understand the extent to which they will bring value to underground mining customers.
Michel Dubois was part of a pillar research center of IVADO, and has now spent the last 7 years at Newtrax in the mining industry. “It’s certainly an advantage to have years of experience on both sides of the equation: researcher in applied mathematics for Artificial Intelligence, and practitioner in underground mining.”
Newtrax is currently focused on an algorithm for vehicle maintenance using the vehicles’ hundreds of sensors to try to predict when components will break or fai. “We have some preliminary results in one mine site in North America and it’s looking very promising.” said Dubois.
Newtrax is helping mines understand their IoT data needs leading the way to a fit-for-purpose ‘Nervous System’ underground that allows for timely and appropriate reaction upon complex real-time events.
With that ‘Nervous system’ producing quality datasets, Newtrax customers can start unlocking unprecedented capabilities to optimize their underground operations.
Newtrax customers data is anonymized, labeled and converted into facts tables and dimensional tables (star schemas, centipede schemas, and more) so that Machine Learning algorithms can consume the information. Newtrax is becoming the expert in quality of data for Machine Learning with a variety of data detection and improvement techniques which are purpose-built for the mining industry and its business cases. Each customer’s data is different. The combination of all these different datasets creates new knowledge that Machine Learning leverages. Value goes back to each customer benefitting from problems being solved at an international level for underground mining.
“Without a doubt, Newtrax technologies, the world leader in big data from underground hard rock mines, holds a unique worldwide position to driving value for its customers through Machine Learning. Newtrax is fortunate to be already working in core operations with leaders in underground mining worldwide, we already have hands-on experience capturing and managing big data from underground mines, and we work with world-class technology and research partners. It’s an honor to be part of the digital revolution in mining.” – concludes Dubois.
Interview with : Michel Dubois, Vice President QA,Newtrax
Mathematician, PhD candidate in Engineering